摘要
针对现有数字调制中多种调制类型混合时的调制识别问题,结合目前对单一调制类型识别各种方法,提出一种综合分类器来完成AWGN环境中的通信信号调制识别。该方法综合使用决策树和RBF神经网络,利用神经网络可以同时对多个特征参数进行非线性优化组合,得到的超曲面能够对整个特征空间进行较精细的分割,从而提高调制识别的整体性能。识别时以决策树为主,以RBF神经网络为辅,其识别效果良好。结果表明,复杂类型调制识别通过合理选择信号特征和分类器,优化识别方法,在一定程度上可以达到相应的识别要求。
According to the modulation recognition when various types of modulation are mixed in the existing digital modulation, combining the various methods of recognizing single modulation type, a method is put forward which utilizes a comprehensive classifier to complete the communication signals modulation recognition in AWGN environment. This method uses decision tree and RBF neural network comprehensively, and can simultaneously carry out the nonlinear optimization re- grouping of multiple characteristic parameters by means of neural network. The super curved surface can split the whole character space subtly, thus the overall capacity of modulation recognition is improved. In the process of recognition, if the decision tree is prior and RBF neural network is complementary, its identification effect is good. The complex type of modula- tion recognition can meet the corresponding identification requirements in some extent by selecting signal characteristics and classifier reasonably, and optimizing the identification method
出处
《现代电子技术》
2012年第12期160-163,共4页
Modern Electronics Technique
基金
四川省教育厅科研项目(11ZB268)
四川文理学院重点项目(2011Z004Z)
关键词
调制识别
似然函数
决策树
神经网络
modulation recognition
likelihood function
decision tree
neural network
